Proliferating cell nuclear antigen (PCNA) immunostaining--a prognostic factor in ovarian cancer?

The measurement of tumour cell proliferation is becoming increasingly recognised in defining prognostic groups. Proliferating cell nuclear antigen (PCNA) immunolocalisation can be used as an index of cell proliferation and may define the extent of departure from normal growth control. The monoclonal antibody PC10 stains PCNA in archival paraffin-embedded tissue. This study investigates its potential as a prognostic marker in early and advanced ovarian cancer. A three-stage immunoperoxidase technique was developed to detect the monoclonal antibody PC10. Archival paraffin-embedded tissue from 19 stage I ovarian tumours (13 malignant and six borderline) and 79 advanced (stage IIb-IV) ovarian tumours (patients entered into the Third North-West Thames Ovarian Cancer Trial) was immunostained with PC10. PC10 immunostaining was performed successfully in 91.8% of cases. The PC10 labelling index (PC10 LI) ranged from 1.5% to 88% with a mean value of 47.4%. Stage I borderline tumours had significantly lower PCNA labelling indexes than stage I malignant tumours (P < 0.048). In advanced disease there was an inverse correlation between PC10 and overall survival, and in those patients who underwent good debulking surgery (37 patients with disease < 2 cm diameter) a low PC10 value (< 36.5%) correlated with improved survival (log-rank trend test for survival, chi 2 = 5.75, P = 0.017). PCNA immunostaining defines a good prognostic subgroup in adequately debulked patients with ovarian cancer. ImagesFigure 1

Recent advances in the study of cell cycle control have suggested the existence of a universal control mechanism common to all eukaryotic cells, regulating the onset of mitosis (Nurse, 1992). Normal cellular growth is controlled by the cell cycle and malignancy arises from derangements of such proliferation.
The extent to which cells escape from normal cell cycle control may reflect their degree of malignancy. It is likely that the association between a high proliferation rate and the degree of tumour invasiveness is a general feature of human solid tumours. However, a high proliferation rate is more likely to be a variable associated with rather than the cause of biological aggressiveness. Departure from normal cell cycle control may be detectable by measurement of abnormalities of the expression of antigens associated with cell cycle control. Such antigens include proliferating cell nuclear antigen (PCNA).
PCNA is an evolutionarily highly conserved acidic protein of 36 kDa, which was independently discovered by Miyachi et al. (1978) as PCNA and by Bravo and Celhs (1980) as cycin. It has more recently been identified as an essential accessory factor to the delta polymerase, which is required for both leading strand DNA replication and DNA repair (Bravo et al., 1987;Prehlich et al., 1987;Toschi and Bravo, 1988;Shivji et al., 1992). It is necessary for DNA replication, DNA repair, cell cycle progression, cellular proliferation and is expressed in late G1-S-phase. PCNA therefore accumulates in cycling cells, and thus in normal tissue PCNA immunolocalisation can be used as an index of the degree of cell proliferation as staining is confined to proliferating cells (Sarraf et al., 1991). In malignant tissue a high level of PCNA immunostaining may identify aggressive tumours and provide a guide to the proliferation rate of the tumour (Hall et al., 1990).
A three-stage immunoperoxidase technique has been used to detect the monoclonal antibody PCIO raised to genetically engineered PCNA in archival paraffin-embedded tissue. This study aimed to determine whether PCIO immunostaining has a role as a prognostic markerdefining patients with either aggressive or fast-proliferating tumours. Ovarian cancer was chosen for study. The prognosis of ovarian cancer has remained largely unchanged in the past decade, since the introduction of platinum-containing regimens. A few clear prognostic factors have been defined, such as stage, grade and residual disease after initial surgery but these are insufficiently discriminating to either define satisfactorily a good prognostic subgroup or select those patients to whom it may be appropriate to offer an additional treatment modality or more intensive treatment regimens.

Patients
Earl) stage Pathological specimens from patients with stage I disease presenting to one consultant (HEL) at Hammersmith Hospital between 1983 and 1991 (13 malignant and six borderline tumours) were studied.
Advanced stage Advanced-disease patients recruited to the Third North-West Thames Ovarian Cancer Trial between 1985 and 1989 were studied. Archival pathological material was obtained on 79 patients of the total 271, material being obtained on all patients presenting to 13 centres of the 43 involved. Treatment consisted initially of five cycles of carboplatin. Patients who showed no evidence of progressive disease then underwent a second-look laparotomy, and responders were randomised to either a further five cycles of carboplatin or whole abdominal radiotherapy (Lambert et al., 1993). The study group is those patients whose tissue was stained with PC1O and was defined on the basis of intention to treat, on recruitment to the trial. The 271 patients entered in the Third North-West Thames Ovarian Cancer Trial are defined as the whole group.

Histological Classification
The histological samples from diagnostic surgical specimens obtained at initial laparotomy were used. Sections were stained with haematoxylin and eosin and graded by one pathologist, on receipt of the slides from the referring centres, according to the degree of cellular differentiation (Decker et al., 1972). PCNA immunostaining A number of monoclonal antibodies to PCNA were raised by Waseem and , and of 11 with anti-PCNA specificity six reacted with formalin-fixed material, of which PC10 had the highest avidity by enzyme-linked immunosorbent assay (ELISA). PC10 has the advantage of recognising PCNA in archival material. Formalin-fixed and paraffin-embedded sections were cut at 4 mm, mounted on poly-L-lysine-coated glass slides and air dried overnight at room temperature. Sections were dewaxed, dehydrated with alcohol and then immersed for 15 min in distilled water with 0.3% hydrogen peroxide to block endogenous peroxidase activity. Immunostaining was performed using the ABC method (Dakopatts UK) with a primary incubation for 12 h at 4°C with PC10 (kindly donated by Professor David Lane, University of Dundee) at a dilution of 1:20 (cell culture supernatant). The second layer consisted of biotinylated goat anti-mouse IgG (Dako, 1:200 dilution) and was applied for 1 h. After rinsing. the final incubation was for 1 h with a streptavidin-biotin-peroxidase complex (Dakopatts, UK).
Diaminobenzidine-hydrogen peroxide was employed as a chromogen and a light Cole's haematoxylin counterstain was used. In negative controls the primary antibody was replaced with phosphate-buffered saline, while human tonsil acted as positive controls.

Cell counting
The slides used for PCNA staining and counting were the consecutive sections after the representative haematoxylin and eosin histologically examined section. Numbers of sections available for staining and counting varied from 1-6 (mean 3). All immunostained sections were examined using a x 40 objective with an eyepiece graticule. Tumour cells but not normal cells were then counted in randomly selected consecutive high-power fields. PCIO immunostaining was assessed in 2000 cells and recorded as a percentage. Scoring of PCIO was carred out without prior knowledge of the histological grade or clinical stage. A cell was considered positive if there was any nuclear staining present.

Mitotic index
The mitotic index was measured as the number of mitotic figures per 2000 tumour cells (%). PCIO labelling index The PCIO labelling index was defined as the number of tumour cells with nuclear PCIO immunostaining divided by the total number of tumour cells and expressed as a percentage: PCIO labelling index = no. of +ve tumour cells 100 (PC1O LI) total no. of cells counted The PCIO labelling index was taken as an approximate measure of proliferative fraction. Sarraf et al. (1991) have shown that, although PC1O overestimates the S-phase fraction, it is a good measure of the growth fraction in many tissues.
Statistical analssis Overall survival, in months, was calculated from the time of initial presentation until death or loss to follow-up and was used to generate Kaplan-Meier log-rank survival curves (Kaplan & Meier, 1958). The data were divided into tertiles and, using these as ordered categories, the log-rank trend test was performed. Univariate analysis of known prognostic factors was performed using the Mann-Whitney test.

Results
Ninety-eight ovanran tumours were immunostained: six borderline stage 1, 13 malignant, stage I; 13 stage II, 55 stage III and 11 stage IV. Nuclear staining was adequate in 91% of cases and scored positive or negative. Ten cases were excluded as no staining was visible. This appeared to be due to technical failure: because of the poor quality of the archival material heat had to be used to assist adherence of the section to the glass slide, which resulted in no staining.
PCNA immunostaining is illustrated in Figure la and b. The characteristics of the study group as compared with the whole group are outlined in Table I. including the proportion of patients in the study group and whole group who were suitable for randomisation and the histological grade and subgroup.
A lack of concordance was seen between mitotic index and PCIO LI. The mitotic index ranged from 0.005% to 5%. Cell cycle times for tumours are thought to be of the order of 8-33 h (Bresciani. 1965;Frankfurt. 1967) and PCNA was thought to be expressed in S-phase and early G1; a ratio of PCIO LI to mitotic index of approximately 10-20 might therefore be expected. The recorded ratios in these patients ranged from 7 to 330, with 75% of patients having a ratio of >20. Such a disproportionately high expression of PCNA may indicate the extent of cell cycle deregulation in those tumours in which PC 10 LI is significantly greater than mitotic index and may be indicative of a biologically aggressive tumour. Mitotic index relates to grade, and PCIO LI was found to be independent of grade.  Advanced disease PCIO LI in advanced disease ranged from 1.5% to 88% with a mean value of 47.4%. Sixty-nine tumour samples were stained successfully with the antibody and suitable for analysis. No selection bias was identified. Of the 16 patients randomised to receive carboplatin. ten completed ten cycles of chemotherapy. On univariate analysis PC1O LI did not correlate with known prognostic factors: stage, grade or residual disease (Mann-Whitney P> 0.05). The survival by stage of the PCIO study group is shown in Figure 2. Overall survival for the study group correlated inversely with PCIO LI (see Figure 3). Of the 37 patients who underwent good debulking surgery (<2cm residual disease), 20 were randomised and received additional therapy (ten to carboplatin and ten to radiotherapy). Of the remaining 17, five achieved complete remission, nine suffered disease progression, two had adhesions preventing assessment and one died during chemotherapy. In this group a low PCIO value (the lower tertile was 0-36.5%) correlated with improved survival (see Figure   4) (log-rank trend test for survival, jx = 5.75, P= 0.017). Figure 5 this excludes the possibility that unrecognised biological factors have confounded the data. There were too few events in stage II and IV patients for a survival curve to have been of value.

Survival by PCIO LI for stage 3 well-debulked patients only is shown in
Rapid tumour growth rate is generally associated with poor prognosis (Tubiana and Courdi, 1989). Derangements of cell cycle control and hence proliferation may result in the development of malignancy, and the degree of malignancy may be related to the extent to which cells escape from this control. Such escape may be detectable by measuring alteration in expression of genes associated with the cell cycle including p53, Rb and the PCNA gene (De Caprio et al., 1989;Hall et al., 1990;Levine et al., 1991).
The PCNA gene is necessary for DNA replication, DNA repair, cell cycle progression and cellular proliferation. Exposure of cells to antisense oligodeoxynucleotides to PCNA results in complete cessation of DNA synthesis and cellular proliferation (Jaskulski et al., 1988;Liu et al., 1989). Detection of PCNA expression in a cell indicates proliferationin normal tissue PCNA staining is confined to proliferating cells. PCNA is expressed in late G,-S-phase, and several models support the hypothesis that late GI events play a major role in the control of cell proliferation. When some GI-arrested, temperature-sensitive mutants of the cell cycle (Burstin et al., 1974;Talavera and Basilico, 1977)   neoplastic transformation; its control may be closely related to that of the PCNA gene. Conditional expression of wildtype p53 protein in a cell line (GM47.23) derived from human glioblastoma multiforme had a negative effect on cell proliferation (Mercer et al., 1990). It has since been shown in this cell line that inhibition of cell cycle progression into S-phase is accompanied by selective down-regulation of PCNA mRNA and protein expression (Mercer et al., 1991). Inactivation of the tumour-suppressor activity of p53 appears to be an almost universal step in the development of human cancers (Hollstein et al., 1991). In colorectal adenomas its overexpression has been correlated with increased proliferative rate, as detected by PCNA immunostaining, and may underlie the dysplasia and loss of proliferative control characteristic of adenomas with high malignant potential (Pignatelli et al., 1992). The relationship between p53 and PCNA has also been demonstrated in prostate carcinoma, in which again p53 staining has been found to correlate with PCNA expression (Visakorpi et al., 1992). Previously measurement of cell proliferation kinetics has included flow cytometric (FCM) analysis of DNA (S-phase fraction) or immunohistochemical detection of bromodeoxyuridine. In ovarian cancer S-phase fraction (SPF) has been found to be of prognostic significance. A number of studies of flow cytometric analysis of SPF in epithelial ovarian cancer have found it a useful and independent prognostic factor (Vohm et al., 1985;Rutgers et al., 1987;Kallioniemi et al., 1988;Barnabei et al., 1990). Ovarian tumours of borderline malignancy (OTBM) have been assessed by means of flow cytometry as part of larger studies including both borderline and malignant neoplasms. In most, SPF has been found to be significantly lower in borderline tumours, in comparison with malignant epithelial ovarian cancer.
A method of measuring tumour cell proliferation which can be performed on archival, paraffin-embedded material has many advantages. Studies on a range of malignancies have shown that PCNA LI correlates with other means of measuring tumour proliferation such as flow cytometnrc analysis of S-phase fraction, tritiated thymidine labelling index (LI), bromodeoxyuridine (BrdU) identified labelling index and Ki67 labelling index (Dawson et al., 1990;Allegranza et al., 1991). In most studies PCNA LI values are higher than tritated thymidine LI and BrdU LI and S-phase fractions calculated from flow cytometric DNA histograms (Garcia et al., 1989). This may, in part, be explained by the fact that PCNA is expressed during GI, S, G2 and M-phases of the cell cycle and not just restricted to S-phase.
Comparisons have been made between PCNA immunostaining and Ki-67 as a means of assessing proliferative activity (Louis et al., 1991;van Dierendock et al., 1991). Not all studies show a correlation between the two methods and a suggested explanation is that PCNA expression may be deregulated in malignancy and expressed persistently in some cells which are not actively dividing (Rosa et al., 1992).
The number of malignancies which have been investigated with PC1O immunostaining is now very large, and in many the results have been compared to established prognostic factors. In gastric carcinoma no correlation was seen between tumour stage, histology or presence of lymph node metastases but, examining survival above and below the median PCIO LI, those with a higher index tended to have a worse prognosis, though this was not statistically significant (Jain et al., 1991). In 178 transitional cell bladder cancers (TCCs) the proportion of PCNA-positive nuclei was related to T status, N status, WHO histological grade, and predicted progression in T, N and M categories. The fraction of PCNA-positive nuclei predicted survival in the entire cohort. In multivariate analysis of the PCNA LI showed independent predictive value as a significant prognostic variable in TCC (Lipponen and Eskelin, 1992). Studies in lymphomas have also demonstrated that PCNA immunostaining may be of use as a marker of proliferative activity, with some prognostic significance (Kamel et al., 1991;Woods et al., 1991). In 194 patients with stage III carcinoma of the cervix treated with radiation therapy alone a strong correlation was found between the PC10 index and prognosis, suggesting its potential as a prognostic indicator in patients with advanced cancer to be treated with this modality (Oka et al., 1992). The vast majority of patients with ovarian cancer present with advanced-stage disease. Adequate debulking surgery is of prognostic importance and, although the disease is chemoresponsive, relapse is often inevitable with a 5 year survival rate in stage III disease of about 20%, virtually unchanged in the last few decades (American Cancer Society, 1986). This study shows that there is a trend towards high PC1O values in patients with poor survival and, importantly, in those patients who had undergone optimal debulking surgery (<2 cm residual disease), it selects out a subgroup who have double the median survival (44 as compared with 22 months). This suggests the need for a prospective randomised trial to establish the clinical significance of this prognostic marker. A knowledge of the proliferative activity of tumours may be useful in the evaluation of prognosis and provide a means of selecting patients for appropriate therapy. Ultimately it may be appropnrate to select patients for more intensive treatment regimensfor instance maintenance cytokine therapy or radiolabelled monoclonal antibody therapyon the basis of the biology and malignant potential of disease as determined by quantitation of proliferation, measured in this way.